Signal-processing device and computer-readable recording medium with signal-processing program recorded thereon

Abstract

A signal-processing device, which processes transparent images each expressed by transparent superimposition of a target such as an affected part and a background, includes a transparent image acquiring unit that acquires the transparent images each including the target at a plurality of times; a calculating unit that calculates a component of the transparent image originating from the target and a component of the transparent image originating from the background at a given time t among the plurality of times in association with estimated values of at least one of a distribution originating from the target and a distribution originating from the background and that evaluates consistency with the transparent images; and an updating unit that updates the estimated values of at least one of a distribution originating from the target and a distribution originating from the background at the time t based on the evaluation result.

The invention claimed is: 
     
       1. A signal-processing device which processes transparent images each expressed by transparent superimposition of a target and a background different from the target, the device comprising:
 at least one processor configured to acquire the transparent images each including the target at a plurality of times, 
 wherein: 
 the at least one processor estimates a first component, which originates from the target, of the transparent image at a given time t among the plurality of times based on the transparent image at the given time t and an estimated value of a second component, which originates from the background, of the transparent image at the given time t, estimates a movement amount of the target based on the estimated first component, and updates the estimated first component and the estimated second component at the given time t based on the estimated movement amount and the acquired transparent images; 
 a portion in each of the transparent images, at which portion the target is superimposed on the background, is a sum of the first component of the transparent image on the portion and the second component of the transparent image on the portion; and 
 both of the first component and the second component in the portion comprise non-zero values for values representing the same image coordinates in an original image. 
 
     
     
       2. The signal-processing device according to  claim 1 , wherein the estimated first component and the estimated second component at the given time t are updated when shifting the time t to another time. 
     
     
       3. The signal-processing device according to  claim 1 ,
 wherein the component includes luminance information indicating luminance intensities. 
 
     
     
       4. The signal-processing device according to  claim 3 , wherein the at least one processor:
 generates a measurement vector b indicating the luminance information in each transparent image at the time based on the transparent images at the plurality of times; 
 sets or updates a target variable vector I a  including the luminance information originating from the target of the transparent image at the time t and a background variable vector I b  including the luminance information originating from the background of the transparent image at the time t; 
 sets or updates a function A expressing position information in the transparent images of the target variable vector I a  and background variable vector I b  at the plurality of times; 
 evaluates consistency among the measurement vector b, a variable vector I including the target variable vector I a  and the background variable vector I b , and the function A; and 
 updates the function A or the variable vector I based on the evaluation result. 
 
     
     
       5. The signal-processing device according to  claim 1 ,
 wherein the at least one processor: 
 performs statistical analysis of the target and the background; and 
 updates the estimated value based on the evaluation result and the statistical analysis result. 
 
     
     
       6. A signal-processing device which processes transparent images expressed by transparent superimposition of L (where, L is a natural number) targets and a background other than the targets, the device comprising:
 at least one processor configured to acquire the transparent image in regard to the targets at a given time t, 
 wherein the at least one processor:
 generates a measurement vector b (t, t 1 , t 2 , . . . , and t K ) (where, K is a natural number) corresponding to (K +1) times which consists of (K+1)×M×N luminance values (where, M and N are natural numbers) in regions of interest of the (K+1) transparent images acquired at the time t and K times t 1 , t 2 , . . . , and t K  prior to the time t, respectively, each of the regions of interest being a M×N region; 
 updates variable vectors (hereinafter, referred to as target variable vectors I a1 , I 2 , . . . , and I aL ) consisting of at least one luminance value originating from the L targets in the region of interest, respectively and a variable vector (hereinafter, referred to as a background variable vector I b ) consisting of at least one luminance value of the background in the region of interest based on the measurement vector b, the background variable vector I b (t p ) and target variable vector I a (t p ) consisting of L variable vectors I a1 (t p ), I a2 (t p ), . . . , and I aL (t p ) in regard to the region of interest of the transparent image at a time t p  prior to the time t; 
 estimates a movement amount vector representing a movement amount of the targets based on the updated target variable vector I a   
 estimates, based on the estimated movement amount vector, a coefficient matrix A of (K+1) MN ×(L+1) MN regarding the positions of the L targets and the background at the time t and the K times t 1 , t 2 , . . . , and t K , respectively; and 
 calculates a variable vector I including the updated target variable vector I a  and the updated background variable vector I b  so that evaluation of an evaluation value PI defined by a function of the measurement vector b, the variable vector I, the estimated coefficient matrix A:
     PI=f  ( A, I ,b )   [Expression 1]
 
 
 
 is increased, where f is a deterministic or stochastic function, 
 wherein the acquisition of the transparent image by the at least one processor, the generation of the measurement vector b by the at least one processor, the updating of the variable vector I by the at least one processor, and the calculation of the variable vector I by the at least one processor are performed by shifting the time, 
 wherein a portion in each of the transparent images, at which portion the target is superimposed on the background, is a sum of a first component, which originates from the target, of the transparent image on the portion and a second component, which originates from the background, of the transparent image on the portion, and 
 wherein both of the first component and the second component in the portion comprise non-zero values for values representing the same image coordinates in an original image. 
 
     
     
       7. The signal-processing device according to  claim 6 , wherein the at least one processor
 sets given initial values in the target variable vector I a  and 
 sets, in the background variable vector I b , a result obtained by subtracting the target variable vector I a  in which the initial value is set from the measurement vector b. 
 
     
     
       8. The signal-processing device according to  claim 6 , wherein the at least one processor updates the variable vector I using the result calculated by the at least one processor. 
     
     
       9. The signal-processing device according to  claim 8 , wherein the at least one processor
 updates the background variable vector I b  by the background variable vector I b  calculated by the at least one processor, 
 updates (L−1) variable vectors I aj  by setting a result obtained by subtracting the updated background variable vector I b  and the (L−1) variable vectors I ai  (where, i=1, 2, . . . , and L, here, i ≠ j and j is a natural number that satisfies “0 <j <(L +1)”) at a previous given time from the measurement vector b, and 
 updates one non-updated variable vector I aJ  by setting a result obtained by subtracting the updated background variable vector I b  and the updated (L−1) variable vectors I aj  (where, j=1, 2, . . . , and L, here, j ≠ J and J is a natural number that satisfies “0 <J <(L+1)”) from the measurement vector b. 
 
     
     
       10. The signal-processing device according to  claim 7 , wherein the at least one processor estimates the coefficient matrix A based on the variable vector I set based on the initial value and the updated variable vector I. 
     
     
       11. The signal-processing device according to  claim 7 , wherein the at least one processor estimates the coefficient matrix A based on the variable vector I before the updating and the variable vector I after the updating. 
     
     
       12. The signal-processing device according to  claim 10 , wherein the at least one processor calculates the variable vector I using a gradient method. 
     
     
       13. The signal-processing device according to  claim 11 , wherein the at least one processor calculates the variable vector I using a gradient method. 
     
     
       14. The signal-processing device according to  claim 6 , further comprising:
 the at least one processor configured to estimate temporal position variations of the targets based on the target variable vector I a  in which an initial value is set by the at least one processor and the target variable vector I a  updated by the at least one processor. 
 
     
     
       15. The signal-processing device according to  claim 6 , further comprising:
 the at least one processor configured to estimate temporal position variations of the targets based on the target variable vector I a  before the updating performed by the at least one processor and the target variable vector I a  after the updating performed by the at least one processor. 
 
     
     
       16. A non-transitory computer-readable recording medium which records a program causing a computer, which processes transparent images each expressed by transparent superimposition of a target and a background different from the target, to realize:
 a transparent image acquiring function of acquiring the transparent images each including the target at a plurality of times; 
 a first estimating function of estimating a first component, which originates from the target, of the transparent image at a given time t among the plurality of times based on the transparent image at the given time t and an estimated value of a second component, which originates from the background, of the transparent image at the given time t; 
 a second estimating function of estimating a movement amount of the target based on the estimated first component; and 
 an updating function of updating the estimated first component and the estimated second component at the given time t based on the time t based on the estimated movement amount and the acquired transparent images, 
 wherein 
 a portion in each of the transparent images, at which portion the target is superimposed on the background, is a sum of the first component of the transparent image on the portion and the second component of the transparent image on the portion; and 
 both of the first component and the second component in the portion comprise non-zero values for values representing the same image coordinates in an original image.